Red vs White leds around cameras....

564 views
Skip to first unread message

hardik

unread,
Jan 7, 2020, 9:09:45 AM1/7/20
to OpenPnP
All of the builds I've seen so far online have either red LED's around the cameras for lighting or white LEDs.
Can someone explain pro's and con's of each setup for OpenPnP?

John Plocher

unread,
Jan 7, 2020, 12:04:55 PM1/7/20
to ope...@googlegroups.com
Red is the complimentary color of green, and juju nozzles are green.   The red light turns the green into a low contrast “grey”, which makes it easy to filter them out of the image when identifying components.

While white leds work, they may take more effort to work well and/or reliably...

There was a long thread here on this topic a month or so ago...

  John

On Tue, Jan 7, 2020 at 6:09 AM hardik <hardikp...@gmail.com> wrote:
All of the builds I've seen so far online have either red LED's around the cameras for lighting or white LEDs.
Can someone explain pro's and con's of each setup for OpenPnP?

--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/e8e742aa-c0a1-4716-99a3-70970cc5d72c%40googlegroups.com.

Al Yard

unread,
Jan 7, 2020, 3:59:49 PM1/7/20
to OpenPnP
I converted my bottom camera to red backlighting a couple of months ago.  I found with the white backlight, the Juki nozzle looked bright enough that it was hard to find a reliable threshold value to differentiate it from the part pins.  With the red backlight, the Juki nozzle appears as a darker grey and is much easier to differentiate from the part pins.

Al


On Tuesday, 7 January 2020 09:04:55 UTC-8, John Plocher wrote:
Red is the complimentary color of green, and juju nozzles are green.   The red light turns the green into a low contrast “grey”, which makes it easy to filter them out of the image when identifying components.

While white leds work, they may take more effort to work well and/or reliably...

There was a long thread here on this topic a month or so ago...

  John
On Tue, Jan 7, 2020 at 6:09 AM hardik <hardikp...@gmail.com> wrote:
All of the builds I've seen so far online have either red LED's around the cameras for lighting or white LEDs.
Can someone explain pro's and con's of each setup for OpenPnP?

--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ope...@googlegroups.com.

Marek T.

unread,
Jan 7, 2020, 4:20:20 PM1/7/20
to OpenPnP
Hi Al,
The LEDs you have used to your convertion were some special type? Or just ordinary red?

Maple_Dude

unread,
Jan 7, 2020, 4:52:37 PM1/7/20
to OpenPnP
I'm glad someone asked this question - I was wondering the same thing about Red vs White LED's. I had tried amber LED's and found that they did not produce the desired effect. 

SMdude

unread,
Jan 7, 2020, 6:15:37 PM1/7/20
to OpenPnP
On my juki it has a red light for the top camera and it picks up fiducials really well. I think the camera itself is black and white, the display is monochrome green. I don't know if the video board does any further processing/ conditioning of the image.
For part alignment it uses laser on the fly.

Pins and pads should show up bright with red light. I am thinking of setting this up on my openpnp machine.

Al Yard

unread,
Jan 7, 2020, 7:09:43 PM1/7/20
to OpenPnP
Marek,

No, nothing special.  Just as an experiment, I took one of the those decorative LED light strips, it has alternating red/green/blue LEDs.  I managed to get enough of the the light strip in my backlight fixture to have 8 red LEDs.  Later, I'm going to do up something better, but even this mickey-mouse effort works quite well.  Unfortunately, I don't know the wavelength of these red LEDs, but I'm guessing anything in the red range is going to be an improvement. 

Al

Marek T.

unread,
Jan 7, 2020, 8:59:30 PM1/7/20
to OpenPnP
I also consider an experiment like that. But first I want to build a coaxial light housing then changing of the light plate will be very simple and good to observe the difference.

Harjit Singh

unread,
Jan 7, 2020, 9:52:46 PM1/7/20
to ope...@googlegroups.com
What are you thinking for the coaxial light housing?

Are you going to use a piece of regular glass or are you going to use a one-way glass?


On Tue, Jan 7, 2020, 5:59 PM Marek T. <marek.tw...@gmail.com> wrote:
I also consider an experiment like that. But first I want to build a coaxial light housing then changing of the light plate will be very simple and good to observe the difference.

--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/db2ba2b4-fbb5-4523-b690-650e415864ed%40googlegroups.com.

Marek T.

unread,
Jan 8, 2020, 3:59:03 AM1/8/20
to OpenPnP
I have a sheet of the mirror glass.

ma...@makr.zone

unread,
Jan 8, 2020, 4:49:50 AM1/8/20
to ope...@googlegroups.com

Hi

For today's standard OpenPNP bottom vision pipeline it does not matter either way.

BUT:

From an information theoretical standpoint and for future improved and/or special case pipelines, with just red LEDs you lose a lot of information. It's like just using the R channel of your RGB camera, you lose two "dimensions" to distinguish object from the background. A bit like closing one of your eyes and seeing things only in 2D instead of 3D.

Here's one application that won't be possible with red LEDs:
https://github.com/openpnp/openpnp/pull/922

It is true, that this feature is only necessary for rare special cases, but you might one day be glad to be able to do it.

This is even more true for the down-looking camera. I plan to release a feeder that extensively uses the green screen effect and this only works with RGB cameras and white lighting.

IMHO, red LEDs only make sense for physically monochrome cameras and monochrome cameras nowadays only makes sense if you have a hardware integrated hi-speed camera, because the sensor pixels are larger and therefore faster, the image data is less and faster to process. But immediately forget this with USB cameras or anything cheap like it, the speed difference will not matter against communication delays etc. and there are many much more promising potentials for speed improvements in OpenPNP.

Personally, I would never go for red LEDs.

_Mark

Marek T.

unread,
Jan 8, 2020, 5:01:51 AM1/8/20
to OpenPnP
Hi Mark,

Have you made some tests how are visible in red and white light different types of fiducials? I mean usually extra flat and perfectly visible ENIG or Organic, against usualy non-flat (little convex) and wild light reflecting in different directions HAL pads.
No difference?

W dniu środa, 8 stycznia 2020 10:49:50 UTC+1 użytkownik ma...@makr.zone napisał:

Hi

For today's standard OpenPNP bottom vision pipeline it does not matter either way.

BUT:

From an information theoretical standpoint and for future improved and/or special case pipelines, with just red LEDs you lose a lot of information. It's like just using the R channel of your RGB camera, you lose two "dimensions" to distinguish object from the background. A bit like closing one of your eyes and seeing things only in 2D instead of 3D.

Here's one application that won't be possible with red LEDs:
https://github.com/openpnp/openpnp/pull/922

It is true, that this feature is only necessary for rare special cases, but you might one day be glad to be able to do it.

This is even more true for the down-looking camera. I plan to release a feeder that extensively uses the green screen effect and this only works with RGB cameras and white lighting.

IMHO, red LEDs only make sense for physically monochrome cameras and monochrome cameras nowadays only makes sense if you have a hardware integrated hi-speed camera, because the sensor pixels are larger and therefore faster, the image data is less and faster to process. But immediately forget this with USB cameras or anything cheap like it, the speed difference will not matter against communication delays etc. and there are many much more promising potentials for speed improvements in OpenPNP.

Personally, I would never go for red LEDs.

_Mark


Am 07.01.2020 um 18:04 schrieb John Plocher:
Red is the complimentary color of green, and juju nozzles are green.   The red light turns the green into a low contrast “grey”, which makes it easy to filter them out of the image when identifying components.

While white leds work, they may take more effort to work well and/or reliably...

There was a long thread here on this topic a month or so ago...

  John
On Tue, Jan 7, 2020 at 6:09 AM hardik <hardikp...@gmail.com> wrote:
All of the builds I've seen so far online have either red LED's around the cameras for lighting or white LEDs.
Can someone explain pro's and con's of each setup for OpenPnP?

--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ope...@googlegroups.com.
--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ope...@googlegroups.com.

Dave B.

unread,
Jan 8, 2020, 8:41:23 AM1/8/20
to ope...@googlegroups.com
Simple solution:  NeoPixel LEDs :)  You can have any color of the rainbow.
-D

To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/16f4bd65-44c5-4d73-8097-bc7e1f3691a0%40googlegroups.com.

Marek T.

unread,
Jan 8, 2020, 8:53:35 AM1/8/20
to OpenPnP
If remember well it was tested by Jason and really not recommended as the source of non-stable (pwm blinking??) light. Maybe I remember wrong...

Jason von Nieda

unread,
Jan 8, 2020, 8:56:13 AM1/8/20
to ope...@googlegroups.com
You remember right. NeoPixel PWM is too slow and causes strobing.

Jason

To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/480bb562-9fde-4fcf-9d99-43ac12cd8048%40googlegroups.com.
--
Sent from my BeOS enabled toaster

E O

unread,
Jan 8, 2020, 10:00:16 AM1/8/20
to OpenPnP
dotstars (APA102) are very similar to neopixels but have a much higher PWM frequency and should be suitable.

Mike M.

unread,
Jan 8, 2020, 10:11:03 AM1/8/20
to OpenPnP
The LuMini line uses the same LED used on our Lumenati boards, the APA102, just in a smaller, 2.0x2.0 mm package. This allows for incredibly tight pixel densities, and thus, a more continuous ring of color. While the LuMini Rings come in different sizes, they all operate in a similar fashion......
Mike

ma...@makr.zone

unread,
Jan 8, 2020, 10:11:57 AM1/8/20
to ope...@googlegroups.com

Hi Marek

the following will sound scoolmasterly, sorry, English is a foreign language and it seems I can't express myself without sounding like that :-}

Remember that what humans call "Color" is governed by what our eyes can perceive. It happens, we have light sensitive receptors for three wavelength ranges, the S, M, L cone cells. While not directly corresponding to the Red, Green, Blue wavelengths it seems our brain can differentiate these three basic colors from the somewhat overlapping cone wavelength sensitivities.

https://en.wikipedia.org/wiki/Color_vision#Physiology_of_color_perception

Btw. Bees see more wavelengths (e.g. ultraviolet), so this is really only valid for our species.

Consequently what we call "white" is a "human perception of sunlight" i.e. the R, G, B components all dialed up. A bee would not call it "white" :-)

If we build color displays and cameras with RGB emitters/sensors, these match that underlying model. But once the devices are built on these assumptions, there is no way around it. All you can do is communicate color in these three channels (I'm assuming affordable consumer products here, but see P.S.).

If you have an RGB camera, it does really make no sense to have red LEDs, because all you need to do to get the same image with white light is to suppress the green and blue channels in the image.

I did that with an image by Mike M.:

The same effect can probably be achieved in a pipeline by using MaskHSV twice (suppress blue and green).

To answer your question: with these devices Fiducial recognition simply cannot be better with colored lighting!

I would instead look at better diffusion of the light.

_Mark

P.S. *) There might be one exception: Infrared. For technical reasons, cameras are usually sensitive to infrared if you remove a filter from it (some can be bought with the filter removed). However I have no idea if infrared would be any better for fiducial recognition. If this were the case, then it might make sense to add infrared LEDs in addition to the white light. Infrared might also be less susceptible to ambient light changes.

To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/16f4bd65-44c5-4d73-8097-bc7e1f3691a0%40googlegroups.com.

Marek T.

unread,
Jan 8, 2020, 10:41:30 AM1/8/20
to OpenPnP
Great lecture as always in your performance :-). No irony just the fact!
But, all you can do in pipeline processing, maskHSV etc, you operate on the white spectrum light reflected from the pad. I was curious whether the "single" red colour is maybe reflecting a bit other than white, and then easier to use.
It's not I'm the fan of the red light or saying you are not right.

I made may type of diffusors, light sourcing (enlighting straightly down, or differentially angled) and finaly best efect got with last Michael G. project. But all this is not it, and nothing is working perfectly with HAL pads when are poorely finished (little convex). Probably only the coaxial light makes real sense, specially after this what I have seen in Essemtec machine.

Infrared sounds interresting. But I heard it works best when the fiducial pad is covered with the solder mask!

Brynn Rogers

unread,
Jan 8, 2020, 11:20:19 AM1/8/20
to OpenPnP

On Wednesday, January 8, 2020 at 9:11:57 AM UTC-6, ma...@makr.zone wrote:
[...]

_Mark

P.S. *) There might be one exception: Infrared. For technical reasons, cameras are usually sensitive to infrared if you remove a filter from it (some can be bought with the filter removed). However I have no idea if infrared would be any better for fiducial recognition. If this were the case, then it might make sense to add infrared LEDs in addition to the white light. Infrared might also be less susceptible to ambient light changes.


Hi Mark, 
    I was at a robot competition once years ago ( a trinity collage firefighting robot contest), and many robots had IR sensors for distance(at that time).    The Gymnasium had skylights, and mercury vapour or hailide lighting, and there were many people taking photographs -  and many of the robots were going nuts because they were being flooded with intense IR from many sources.

   My point is that while IR might or might not be useful for machine vision, it will be just as susceptible to ambient lighting conditions as visible light sensors.   Also if you stick to visible light our eyes would be able to give us some Idea of the ambient levels we are dealing with, while with IR it is difficult to tell what ambient conditions might be like. 

Brynn

Mike M.

unread,
Jan 8, 2020, 11:26:20 AM1/8/20
to OpenPnP
Marek can you post un image from your fiducial ?
Mike

Mike M.

unread,
Jan 8, 2020, 11:53:35 AM1/8/20
to OpenPnP
Hi all,
With Red downligh - I post enclosed what I see with Red Coaxsial Down light on one of my PCBs - original and as per Camera Screen shot fid 1 + 2 enclosed.
I hope this helps - What is the Question Marek :-)(  !?

Mike
DOWN_2020-01-08_17.40.13.224.png
DOWN_2020-01-08_17.38.46.785.png
PCB_42mm.jpg

John Plocher

unread,
Jan 8, 2020, 12:07:24 PM1/8/20
to ope...@googlegroups.com
> Simple solution:  NeoPixel LEDs :)  You can have any color

Nope.

While your eyes may perceive the mix as any color,  on a spectrum analyzer the light is simply a trio of independent monochromatic peaks for R, G, and B at whatever wavelength each of the leds is emitting.

  John

Greg Wroblewski

unread,
Jan 8, 2020, 1:20:25 PM1/8/20
to OpenPnP
I noticed that on my PnP machine (a Chinese build), the LEDs appear to be in the "deep red" range (longer wavelengths), yet the image comes up very crisp and clear. It looks like the cameras have the infrared filtering absent or modified. The image processing results are amazing, much better than what I was able to accomplish with white light and standard OpenPNP pipelines. Their algorithms seem to be more sophisticated though as well.

Greg

Mike M.

unread,
Jan 8, 2020, 4:06:45 PM1/8/20
to OpenPnP
Greg - can you post some of your images ?
Thanks
Mike

Greg Wroblewski

unread,
Jan 9, 2020, 3:23:37 AM1/9/20
to OpenPnP
I assume you mean the "red" ones. The machine has four fast analog cameras and one USB high resolution camera. I only run the original software on it so far (the control protocol is known), so I would have to figure out how to get the captures.

Greg


--
You received this message because you are subscribed to a topic in the Google Groups "OpenPnP" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/openpnp/zMPsM9u6KGQ/unsubscribe.
To unsubscribe from this group and all its topics, send an email to openpnp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/d6501e57-4c02-479b-9904-e79591dce16b%40googlegroups.com.

Kylra

unread,
Jan 10, 2020, 4:27:18 AM1/10/20
to OpenPnP
I think this graph are usefull. All metalls have a high reflectance in the IR spectrum.

For top camera we use white light, for bottom camera red light.

Kylra


Sandra Carroll

unread,
Jan 10, 2020, 8:06:47 AM1/10/20
to ope...@googlegroups.com
True all metals do but cameras have their own spectral response which many don’t publish unless you can find the actual sensor data sheet

Most though are weak in the blue and stronger in the green thru IR.  Which is why the IR cut filters are needed


Sent from my iPhone 7 Plus

On Jan 10, 2020, at 4:27 AM, 'Kylra' via OpenPnP <ope...@googlegroups.com> wrote:


--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/2645b570-58a3-47b5-8c8d-79c04b8348ca%40googlegroups.com.

ma...@makr.zone

unread,
Jan 10, 2020, 9:05:28 AM1/10/20
to OpenPnP
Interesting...

A step back: With HASL you get a slightly convex surface (created by surface tension) so it acts like a miniature convex traffic mirror, therefore it tends to reflect stuff all around, including the (usually black) camera pointing at it, the LEDs etc.

File:Traffic Mirror at Curve.JPG

With infrared, this does not change it seems, so we would still need a diffuser, right?

It seems nothing is gained ... but much is lost, because we cannot judge an IR diffuser by our own eyes, can we?

_Mark

Marek T.

unread,
Jan 10, 2020, 9:31:27 AM1/10/20
to OpenPnP
So however you turn, your butt is still at the back?
(except for the concentric light that I still want to believe).

Jim

unread,
Jan 10, 2020, 10:28:57 AM1/10/20
to OpenPnP
I played i little bit with a mirror light (CA-DXW3) and see homogen light only inside the core zone.
The core zone equals the the mirror size (@t 45 deg).

Outside of the core zone HAL fids have the same bad reflectivty as with direct light.
That means that for bigger objects (large chips @ bottom vision) a bigger mirror is needed.

For top vision i prefer white light because it is more "wysiwyg".
And as long as there is no affordable high speed bottom vision system on the market (monochrome because of performance reasons) i stay with usb2/3 color cameras.

ma...@makr.zone

unread,
Jan 10, 2020, 10:45:29 AM1/10/20
to OpenPnP
On Friday, January 10, 2020 at 3:31:27 PM UTC+1, Marek T. wrote:
So however you turn, your butt is still at the back?

Yes, exactly :-D
 
(except for the concentric light that I still want to believe).

 

The concentric light is the best solution, of course. But a good diffuser could get close. The idea is to make the black camera "blind spot" as small as possible, so it will be small in the reflection.


Actually I started some experiments with 3D-printed transparent PETG diffuser.


Imagine a LED ring behind it. Perhaps with some 90°-LEDs on the inner ring edge also lighting the center.


As you see, the diffusor is quite "all around" or you could say "dome like" and the blind spot is small in deed, the chances of it reflecting in the object/fiducial in a disruptive way are reduced (mehopes). 


The 3D-printing layers result in a "natural" diffusion characteristic, the transparent PETG fibers act a bit like light conductors. It works very well with simulated LEDs behind it, quite even-looking. But I haven't had the time to integrate it into a mounting solution on my machine and do real camera shots yet.


It's a fully parametric OpenSCAD file, you can enter the camera and lens data and it will calculate the perfect cone angle for the funnel. The funnel is just outside the FOV cone. Preset for my ELP 6mm lens cameras.


If somebody is interested in this early untested stage, I can provide the file.


_Mark

Marek T.

unread,
Jan 10, 2020, 11:21:00 AM1/10/20
to OpenPnP
 @Jim, are you able to upload some image of as poor HAL fidicial as you can find, and positioned exactly in the centre of FOV?
And I assume you don't have any cam transformations applied to the camera settings. If it is, an image is useless (telling not much).

@ Mark, I feel confused or worried about my "English" (I still haven't decided what more) :-(
"Imagine a LED ring behind it. Perhaps with some 90°-LEDs on the inner ring edge also lighting the center."
So <behind> ( it sounds like meaning on the outer surface of your cloche) or on the <inner ring egde> (but which ring?) ?


Kylra

unread,
Jan 10, 2020, 11:45:00 AM1/10/20
to OpenPnP
It would look like this (bottom cam with yellow and red 120°LEDs).
Transparent PETG isn't diffuse. Amolen PETG diffuse transparent is a better option.

Kylra

botcam2.jpg

botcam.jpg

ma...@makr.zone

unread,
Jan 10, 2020, 11:46:58 AM1/10/20
to ope...@googlegroups.com

Don't laugh, that's just hand-doodled :-D

--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.

Marek T.

unread,
Jan 10, 2020, 11:57:17 AM1/10/20
to OpenPnP
aaaaa, no laugh, clear now :-).


W dniu piątek, 10 stycznia 2020 17:46:58 UTC+1 użytkownik ma...@makr.zone napisał:

Don't laugh, that's just hand-doodled :-D


Am 10.01.2020 um 17:21 schrieb Marek T.:
 @Jim, are you able to upload some image of as poor HAL fidicial as you can find, and positioned exactly in the centre of FOV?
And I assume you don't have any cam transformations applied to the camera settings. If it is, an image is useless (telling not much).

@ Mark, I feel confused or worried about my "English" (I still haven't decided what more) :-(
"Imagine a LED ring behind it. Perhaps with some 90°-LEDs on the inner ring edge also lighting the center."
So <behind> ( it sounds like meaning on the outer surface of your cloche) or on the <inner ring egde> (but which ring?) ?


--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ope...@googlegroups.com.

ma...@makr.zone

unread,
Jan 10, 2020, 12:12:01 PM1/10/20
to OpenPnP
Cool, Kylra.

The diffusion is not created by the PETG but rather by the print layers' coarseness. It works more like the "antithesis" to a Fresnel lens, a bit like the diffuser on a flash light (looks similar).


I assume the layer coarseness is somewhat dependent of the cone angle and the layer height (I used the coarsest available, I think 0.3mm), layer number. Also, the distance to the LEDs is relevant. Yours is quite shallow and seems extremely thin (guess you printed it sideways), the result is very transparent. I since printed a lamp shade and it is completely diffuse.

The cone seemed to work very well in my simulated tests, but I can't guarantee it will in a real implementation. 

_Mark

Kylra

unread,
Jan 10, 2020, 12:16:55 PM1/10/20
to OpenPnP
A funnel with lights need the world ;D


Am Freitag, 10. Januar 2020 17:46:58 UTC+1 schrieb ma...@makr.zone:

Don't laugh, that's just hand-doodled :-D


Am 10.01.2020 um 17:21 schrieb Marek T.:
 @Jim, are you able to upload some image of as poor HAL fidicial as you can find, and positioned exactly in the centre of FOV?
And I assume you don't have any cam transformations applied to the camera settings. If it is, an image is useless (telling not much).

@ Mark, I feel confused or worried about my "English" (I still haven't decided what more) :-(
"Imagine a LED ring behind it. Perhaps with some 90°-LEDs on the inner ring edge also lighting the center."
So <behind> ( it sounds like meaning on the outer surface of your cloche) or on the <inner ring egde> (but which ring?) ?


--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ope...@googlegroups.com.

Kylra

unread,
Jan 10, 2020, 12:35:41 PM1/10/20
to OpenPnP
Hi Mark, I mean something different.
Both are transparent PETG, left is diffuse, right clear.

Kylra

diffuse transparent.jpg

ma...@makr.zone

unread,
Jan 10, 2020, 1:07:35 PM1/10/20
to OpenPnP
Yeah, I understood that. It's another solution and I will keep it in mind, if the coarseness doesn't work per se. I guess the murkiness will also swallow some light, so I'd like to try without first.

Have you printed your diffuser with the diffuse one?

_Mark

Kylra

unread,
Jan 10, 2020, 1:22:26 PM1/10/20
to OpenPnP
No, it's clear.

Jim

unread,
Jan 10, 2020, 1:28:14 PM1/10/20
to OpenPnP
On Friday, January 10, 2020 at 5:21:00 PM UTC+1, Marek T. wrote:
 @Jim, are you able to upload some image of as poor HAL fidicial as you can find, and positioned exactly in the centre of FOV?
And I assume you don't have any cam transformations applied to the camera settings. If it is, an image is useless (telling not much).


@Marek, i'm still waiting for the appropriate lenses.
And because of 1.rule (never change a running system) i want to make sure that this small coaxlight for top vision will work reliable before i take machine apart.

Here some raw images (only mask, no transform applied) from typical hal boards made with current direct light.

fidloc_original_5003976971675844135.png


fidloc_original_2403309266133408701.png


I know pipelining for this is easy, but i think with coaxlight not so much averaging is necessare (i use 5 times) to be on 100% safe side.

Sandra Carroll

unread,
Jan 10, 2020, 1:36:59 PM1/10/20
to OpenPnP
Has anyone looked at use of UV/IR cut filters then a color bandpass filter?
That would cut out allot of un-needed light getting to the camera.

Dave McGuire

unread,
Jan 10, 2020, 3:31:04 PM1/10/20
to ope...@googlegroups.com
On 1/10/20 10:45 AM, ma...@makr.zone wrote:
> Actually I started some experiments with 3D-printed transparent PETG
> diffuser.
...
> It's a fully parametric OpenSCAD file, you can enter the camera and lens
> data and it will calculate the perfect cone angle for the funnel. The
> funnel is just outside the FOV cone. Preset for my ELP 6mm lens cameras.
>
> If somebody is interested in this early untested stage, I can provide
> the file.

Nice, I'd be interested in giving that a try...can you squirt me that
OpenSCAD file?

Thanks,
-Dave

--
Dave McGuire
McGuire Scientific Services, LLC
New Kensington, PA

ma...@makr.zone

unread,
Jan 10, 2020, 4:10:45 PM1/10/20
to OpenPnP
Here it is.
_Mark
CameraDiffuser_v6.scad

Jim

unread,
Jan 10, 2020, 4:44:47 PM1/10/20
to OpenPnP
On Friday, January 10, 2020 at 7:36:59 PM UTC+1, Sandra Carroll wrote:
Has anyone looked at use of UV/IR cut filters then a color bandpass filter?
That would cut out allot of un-needed light getting to the camera.

Most of the M12 lenses i tried so far have a IR-filter applied.

But i think that UV (<380nm) and IR (>780nm) emission is not very high with led lights.
Here the diagram for the leds i use (4000K):

emission.PNG



Marek T.

unread,
Jan 11, 2020, 9:07:15 AM1/11/20
to OpenPnP
@Jim to be sure, these pictures that you've uploaded are made with coaxial light (CA-DXW3) or other? Asking because you've writtten "with current direct light" and I'm not sure what is this so. Frankly speaking I was prepared for something looking better (however, with my light maybe it would be much more bad).

Jim

unread,
Jan 11, 2020, 9:50:05 AM1/11/20
to OpenPnP
Marek, pictures from above are made with my current setup on machine (self-made lightning with simple led ring with primitive diffuser) and as you can see the fids look no very good.

With the coaxial-light i could not make real tests because i did not have appropriate lenses yet :-(

coaxial-light.jpg


Marek T.

unread,
Jan 11, 2020, 10:04:47 AM1/11/20
to OpenPnP
Ok, it explains everything. I have similiar quality with my light. I use re-designed to my machine the project of Michael.

my 3DCNCPNP

unread,
Jan 11, 2020, 11:46:38 AM1/11/20
to OpenPnP
Marek.   Attached is Hot Air Leveled     fiducial pipeline debug output with my Keyence CA-DXR3. (bought used from ebay)
My ELP cam is a little out of focus.
I put a 1206 for size reference.
I have tried a few different lighting setups and this is the best one so far.
I used to polish my fiducials with a white Staedtler eraser and that worked quite well even with bad lighting.

Peter
fidloc_original_8772250110847583335.png

Marek T.

unread,
Jan 11, 2020, 11:50:46 AM1/11/20
to OpenPnP
Hi Peter,

Yes, out of focus but the surface of the fidu looks like dreaming - if this is HAL not some ENIG or Organic, my target!
Thx for this image.

Marek T.

unread,
Jan 11, 2020, 11:51:31 AM1/11/20
to OpenPnP
aa, Hot Air Leveled :-). Perfect so.

Mike M.

unread,
Jan 11, 2020, 12:36:45 PM1/11/20
to OpenPnP
Based on my experiance sometimes it helps that image is a bit out of focus - than for vision you have less adjustments to make for pipline to get out right shape of element you are katching.
Mike

Jason von Nieda

unread,
Jan 11, 2020, 2:51:22 PM1/11/20
to ope...@googlegroups.com
Instead of an out of focus image, add a Gaussian blur stage to your pipeline. Does the same thing and is very common in vision operations.

Jason


--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/e1ecf8c7-ad95-46d9-8655-08b0b8f49e02%40googlegroups.com.
--
Sent from my BeOS enabled toaster

Mike M.

unread,
Jan 11, 2020, 5:11:03 PM1/11/20
to OpenPnP
Yes Jason -quite often in use for a long time now - but first steps were not so easy as it is now form, I was suffering a lot on the beginning :-)
Thanks
Mike

On Saturday, 11 January 2020 20:51:22 UTC+1, Jason von Nieda wrote:
Instead of an out of focus image, add a Gaussian blur stage to your pipeline. Does the same thing and is very common in vision operations.

Jason

On Sat, Jan 11, 2020 at 11:36 AM Mike M. <mike...@gmail.com> wrote:
Based on my experiance sometimes it helps that image is a bit out of focus - than for vision you have less adjustments to make for pipline to get out right shape of element you are katching.
Mike

On Saturday, 11 January 2020 17:46:38 UTC+1, my 3DCNCPNP wrote:
Marek.   Attached is Hot Air Leveled     fiducial pipeline debug output with my Keyence CA-DXR3. (bought used from ebay)
My ELP cam is a little out of focus.
I put a 1206 for size reference.
I have tried a few different lighting setups and this is the best one so far.
I used to polish my fiducials with a white Staedtler eraser and that worked quite well even with bad lighting.

Peter

--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ope...@googlegroups.com.

Mike M.

unread,
Jan 11, 2020, 5:16:31 PM1/11/20
to OpenPnP
Jason
Could you please up date instruction here
of: how to link in vision script & which-where folders should be made and where to store the images for vision to catch them when needed. 
Thanks
Mike

Brynn Rogers

unread,
Jan 11, 2020, 5:30:33 PM1/11/20
to OpenPnP
So if the main problem with imaging fiducials is the reflectiveness of the HASL(solder) or ENIG(gold),    why don't we make the fiducial a ring of metal around a circle of bare board (no soldermask or silk)?
  Wouldn't we then be looking for a dark circle in the middle of the bright metal ring, which maybe is easy for the image pipeline to make out?

Brynn 

Marek T.

unread,
Jan 11, 2020, 5:42:12 PM1/11/20
to OpenPnP
Many times, most of the times, the boards come from different customers with the fiducials they are used to use..

Jason von Nieda

unread,
Jan 11, 2020, 6:35:34 PM1/11/20
to ope...@googlegroups.com
Hi Mike,

I have never used it - it was added via a pull request and I don't have any experience with it. Perhaps, if you are interested in using it, you could figure it out and then update the docs?

Jason


To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/84c3c1d4-e1e5-4d09-9f3d-2e1aef38e973%40googlegroups.com.

Marek T.

unread,
Jan 12, 2020, 7:25:45 AM1/12/20
to OpenPnP
Gents, have you seen this old project?
https://firepickdelta.dozuki.com/Guide/Assembling+the+Up-looking+camera/16
Considering to print it using mentioned diffusing PETG...

Michael Anton

unread,
Jan 12, 2020, 6:32:55 PM1/12/20
to OpenPnP
I have, in fact I still have one of those...  The project is very dead however.  If you want to see the details, then check out the google group: https://groups.google.com/forum/#!forum/firepick.

Marek T.

unread,
Jan 13, 2020, 7:40:11 AM1/13/20
to OpenPnP
Hi Michael,
Am I blind that can't see any files like stl or so available to download for this? Was this shared, do you know?
Images on the forum I found look quite promising.

my 3DCNCPNP

unread,
Jan 13, 2020, 1:08:05 PM1/13/20
to OpenPnP

Marek T.

unread,
Jan 13, 2020, 1:42:20 PM1/13/20
to OpenPnP
Wow! So I will look arround for this diffusing PETG :-).

I have found something like this, coaxial light housing. However not very perfectly designed:
coax-mirror-camera.stl
coax-lighting.stl

Dave McGuire

unread,
Jan 13, 2020, 2:17:43 PM1/13/20
to OpenPnP

Thank you Mark!

-Dave
> --
> You received this message because you are subscribed to the Google
> Groups "OpenPnP" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to openpnp+u...@googlegroups.com
> <mailto:openpnp+u...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/openpnp/bf5c9095-335a-4007-a850-3ed91dd33a7e%40googlegroups.com
> <https://groups.google.com/d/msgid/openpnp/bf5c9095-335a-4007-a850-3ed91dd33a7e%40googlegroups.com?utm_medium=email&utm_source=footer>.
Reply all
Reply to author
Forward
0 new messages